This repo builds offical Instruqt K3s images, based on the official k3s releases.

This runs a daily build, to check for new releases (excluding pre-releases), and builds a new image when there is a new version available.

When run, the version.sh script checks to see if Instruqt has the latest K3S image in GCP. If not, it uses Packer to build it.

Build the container locally

docker build . -t packer-k3s

Run the container locally:

docker run \
  --volume $HOME/.config/gcloud/:/.config/gcloud/ \
  --env GOOGLE_APPLICATION_CREDENTIALS=/.config/gcloud/application_default_credentials.json \
  packer-k3s

To debug image building issues:

docker run \
  --volume $HOME/.config/gcloud/:/.config/gcloud/ \
  --env GOOGLE_APPLICATION_CREDENTIALS=/.config/gcloud/application_default_credentials.json \
  -it --entrypoint ash packer-k3s

NOTE: Google auth credentials are required to allow the container to query for the existing images. To provide local credentials, you may have to run:

gcloud auth login --update-adc

The official Instruqt K3S image includes:

• K3s

• Helm

• Kubernetes dashboard

  • To update to the latest dashboard:

    GITHUB_URL=https://github.com/kubernetes/dashboard/releases
    VERSION_KUBE_DASHBOARD=$(curl -w '%{url_effective}' -I -L -s -S "${GITHUB_URL}/latest" -o /dev/null | sed -e 's|.*/||')
    curl -sL -o files/dashboard.yml "https://raw.githubusercontent.com/kubernetes/dashboard/${VERSION_KUBE_DASHBOARD}/aio/deploy/recommended.yaml"
    # Enable skipping login page
    sed -i '/--namespace=kubernetes-dashboard/a \            - --enable-skip-login' files/dashboard.yml
    

• Systemd service for kubectl proxy

• Bash script to wait for Kubernetes dashboard to start and print the dashboard token (/usr/bin/start.sh)

• Kubectl completion (see section on how to enable kubectl autocompletion)

Images are built upon releases on the K3s repo. This is a list of images available:

instruqt/k3s-v1-29-0

instruqt/k3s-v1-28-5

instruqt/k3s-v1-27-1

instruqt/k3s-v1-26-4

We advise you to use machine type n1-standard-2 or higher to ensure stability.

Use the following config in your Instruqt config.yml to use this image:

version: "2"
virtualmachines:
  - name: kubernetes
    image: instruqt/k3s-v1-27-1
    shell: /usr/bin/start.sh
    machine_type: n1-standard-2

Use the following config in your Instruqt track:

This image supports creating multi-node cluster. Multi-node clusters contain 2 types of machines:

• One Control Plane server
• Zero or more Worker nodes

The configuration for the Control Plane server is exactly the same as for a single node setup. To add a worker node to this cluster:

1. Add another virtual machine, using the same image (e.g. instruqt/k3s-v1-27-1)
2. On that VM, add an environment variable K3S_CONTROL_PLANE_SERVER. The value must be the hostname of the Control Plane server.

This will switch it's runtime mode to Worker, and will join the cluster defined by the Control Plane server.

version: "2"
virtualmachines:
  - name: server
    image: instruqt/k3s-v1-27-1
    shell: /usr/bin/start.sh
    machine_type: n1-standard-2
  - name: worker1
    image: instruqt/k3s-v1-27-1
    shell: /bin/bash
    machine_type: n1-standard-2
    environment:
      K3S_CONTROL_PLANE_HOSTNAME: server
  - name: worker2
    image: instruqt/k3s-v1-27-1
    shell: /bin/bash
    machine_type: n1-standard-2
    environment:
      K3S_CONTROL_PLANE_HOSTNAME: server

To enable kubectl autocompletion, add the following lines to the setup script of your first challenge:

#!/bin/bash
until [ -f /opt/instruqt/bootstrap/host-bootstrap-completed ]
do
    sleep 1
done

echo "source /usr/share/bash-completion/bash_completion" >> /root/.bashrc
echo "complete -F __start_kubectl k" >> /root/.bashrc